A rule-based expert system for automatic implementation of somatic variant clinical interpretation guidelines

Abstract

Precision oncology aims at integrating molecular data into clinical decision making, in order to provide the most suitable therapy and follow-up according to patient’s specific characteristics. A critical step towards this goal is the interpretation of genomic variants, whose presence can be revealed by next generation sequencing. In particular, cancer variant interpretation defines whether the patient harbors genomic alterations that could be targeted by specific drugs, or that were observed as prognostic biomarkers. To standardize somatic interpretation, in 2017 guidelines have been proposed by a working group of associations, including the American Society of Clinical Oncology (ASCO). Automatic tools implementing such guidelines to ease their actual application in the clinical routine are needed. We developed a Rule-based Expert System (ES) that automatically implements ASCO guidelines. ES is an Artificial Intelligence system able to reason over a set of rules and to perform classification, thus emulating human reasoning process. First, we developed automatic pipelines to extract information of over 1500 known diagnostic/prognostic/diagnostic biomarkers from six public databases, including COSMIC and CiVIC. The collected knowledge base is structured in an object-oriented model and the ES is implemented in a Python program through the PyKnow library.